humberto arce et al- triggered alternans in an ionic model of ischemic cardiac ventricular muscle
TRANSCRIPT
-
8/3/2019 Humberto Arce et al- Triggered alternans in an ionic model of ischemic cardiac ventricular muscle
1/12
Triggered alternans in an ionic model of ischemic cardiacventricular muscle
Humberto Arcea) and Alejandro LopezDepartamento de Fsica, Facultad de Ciencias, Universidad Nacional Autonoma de Mexico,Apartado Postal 70-542, 04510 Mexico, Distrito Federal, Mexico
Michael R. GuevaraDepartment of Physiology and Centre for Nonlinear Dynamics in Physiology and Medicine,
McGill University, 3655 Sir William Osler Promenade, Montreal, Quebec H3G 1Y6, Canada
Received 4 March 2002; accepted 19 June 2002; published 23 August 2002
It has been known for several decades that electrical alternans occurs during myocardial ischemia in
both clinical and experimental work. There are a few reports showing that this alternans can be
triggered into existence by a premature ventricular contraction. Detriggering of alternans by a
premature ventricular contraction, as well as pause-induced triggering and detriggering, have also
been reported. We conduct a search for triggered alternans in an ionic model of ischemic ventricular
muscle in which alternans has been described recently: a one-dimensional cable of length 3 cm,
containing a central ischemic zone 1 cm long, with 1 cm segments of normal i.e., nonischemic
tissue at each end. We use a modified form of the LuoRudy Circ. Res. 68, 15011526 1991
ionic model to represent the ventricular tissue, modeling the effect of ischemia by raising the
external potassium ion concentration ( Ko) in the central ischemic zone. As Ko is increased
at a fixed pacing cycle length of 400 ms, there is first a transition from 1:1 rhythm to alternans or2:2 rhythm, and then a transition from 2:2 rhythm to 2:1 block. There is a range of Ko over
which there is coexistence of 1:1 and 2:2 rhythms, so that dropping a stimulus from the periodic
drive train during 1:1 rhythm can result in the conversion of 1:1 to 2:2 rhythm. Within the bistable
range, the reverse transition from 2:2 to 1:1 rhythm can be produced by injection of a well-timed
extrastimulus. Using a stimulation protocol involving delivery of pre- and post-mature stimuli, we
derive a one-dimensional map that captures the salient features of the results of the cable
simulations, i.e., the 1:12:22:1 transitions with 1: 12: 2 bistability. This map uses anew index of the global activity in the cable, the normalized voltage integral. Finally, we put forth
a simple piecewise linear map that replicates the 1:12:2 bistability observed in the cablesimulations and in the normalized voltage integral map. 2002 American Institute of Physics.
DOI: 10.1063/1.1499275
Heart attack is a leading cause of death. Death during a
heart attack is often due to a disturbance in the rhythm
of the heartbeat cardiac arrhythmia. Arrhythmias
arise in this context because of the interruption in the
flow of blood to the heart muscle myocardial is-
chemia. It has been known for several decades that a
beat-to-beat alternation in the pulse or the electrocardio-
gram alternans can be seen at the onset of ischemia,
just before the arrhythmias first start to appear. While
there has been much speculation about a cause-and-effect
relationship between the alternans and the arrhythmias,
there is as yet no firm evidence for this hypothesis. There
is also some evidence that alternans can start up during
ischemia immediately following the occurrence of a pre-
mature ventricular beat triggered alternans. Again,
the exact nature of any causal connection between the
premature beat and the triggering of the alternans is un-
certain. We therefore undertook to see whether we could
shed some light on triggered alternans by trying to elicit
it in an ionic model of ischemic ventricular muscle.
I. INTRODUCTION
Electrical alternans is a cardiac arrhythmia in which
there is a beat-to-beat alternation in the shape of one or more
of the electrocardiographic complexes. Alternans rhythms in
which there is frank alternation of some component of the
electrocardiogram or electrogram e.g., ST-segment, T-wave,ventricular gradient commonly occur during acute myocar-
dial ischemia in both clinical see Refs. 213 in Ref. 1 and
experimental2 9 work. At times, the alternans can be so small
in magnitude that it is not visible to the naked eye, and signal
processing techniques e.g., power spectrum,1012 complex
demodulation,13 or KarhunenLoevre decomposition14 must
be used to establish its existence. There has been much re-
cent interest in detecting this occult alternans to identify pa-
tients at risk of arrhythmia in ischemia as well as in other
situations.1016
aAddress for reprints: H. Arce, Departamento de Fsica, Facultad de Cien-
cias, Universidad Nacional Autonoma de Mexico, Apartado Postal 70-542,
04510 Mexico, Distrito Fede ral, Mexico. Electronic mail:
CHAOS VOLUME 12, NUMBER 3 SEPTEMBER 2002
8071054-1500/2002/12(3)/807/12/$19.00 2002 American Institute of Physics
Downloaded 12 Sep 2002 to 132.248.28.201. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/chaos/chocr.jsp
-
8/3/2019 Humberto Arce et al- Triggered alternans in an ionic model of ischemic cardiac ventricular muscle
2/12
During experimental work on acute coronary ischemia,
the appearance of an alternans rhythm can sometimes be im-
mediately preceded by a spontaneous premature ventricular
contraction PVC.2,4,8,9 One might reasonably think that this
finding is simply a coincidencei.e., that the PVC just hap-
pened to come along just by chance at a time when alternans
would have started up spontaneously anyway. However, the
onset of alternans immediately following a PVC was seen
tens of times in one study,8
in 21 of 31 experiments in an-other study,4 and in 4 out of 8 animals in yet another study.2
There is thus certainly a causal relationship between the PVC
and the alternans: hence the term triggered alternans, 8
which we adopt. Alternans can also cease immediately fol-
lowing a spontaneous PVC,2,4 following an externally in-
duced extrasystole,8 or following a deliberate pause in
stimulation.8 We refer to this phenomenon as detriggering.
It has been known for some time that alternans can exist
in situations where inhomogeneity does not play a role in its
induction: e.g., single ventricular cells,1721 ionic models of
space-clamped ventricular membrane,17,20,2226 ionic models
of homogeneous one-dimensional strands2730 and
2-dimensional sheets29,31 of ventricular tissue, a coupled-maplattice,30 and simple one-dimensional finite-difference equa-
tions stemming from experimental3235 and modeling22,27
work. Alternans has only more recently been seen in models
of inhomogeneous ventricular muscle.1,29,31,36 We thus de-
cided to investigate whether triggered alternans could also be
seen in a model of ischemic ventricular muscle.
II. METHODS
We study an ionic model of a one-dimensional strand of
normal ventricular myocardium, with an area of elevatedKo embedded within its interior to represent the ischemic
zone.1,3743 We use a first-generation model of the ven-
tricular membrane ionic currents, that due to Luo and
Rudy,44 avoiding use of one of the more recent second-
generation models.
In their original formulation, second-generation models
of space-clamped cardiac membrane suffer from two defi-
ciencies: i degeneracy, with nonuniqueness of equilibria
steady-states and limit cycles,4548 and ii very slow long-
term drifts in the variables.41,47,4952 The degeneracy can be
removed by reformulating these models as differential-
algebraic systems,4548,52 rather than as fully-differential sys-
tems, which is the way they were originally formulated. Inone model of spontaneous activity of the sinus node, in
which external stimuli were not delivered, the strategy of
employing the differential-algebraic formulation also abol-
ished drift.47 Drift in another unstimulated sinus node model
was removed by exquisitely adjusting two parameters, main-
taining the original fully-differential formulation.49 In a
model of quiescent ventricular muscle, it has been shown
that the drift is abolished in both the fully-differential and the
algebraic-differential formulations, provided that one takes
into account the ion species injected by the constant-current
stimulus pulses see Fig. 5 of Ref. 52. However, it is not
clear from this paper whether drift was originally present in
the differential-algebraic case before stimulus-current track-
ing was incorporated. This procedure involving keeping
track of the stimulus current is obviously incapable of re-
moving the drifts seen in the fully-differential formulation of
spontaneously active models,4951 where no stimuli are in-
jected. If the fully-differential formulation is used, one is still
left with a system that is degenerate, even if one keeps track
of the stimulus current see, e.g., Fig. 6A of Ref. 52. Finally,
as previously noted,52
should one adopt stimulus-currenttracking for a cable simulation using the fully-differential
formulation, or even using the differential-algebraic formal-
ism if stimulus-current tracking is indeed needed to remove
drift in the space-clamped case, a term to account for inter-
cellular diffusion of the injected ion species and for other
ionic species might then have to be added to avoid drift.
Given all the above uncertainties and complications, we de-
cided to use a first-generation model. We select the Luo
Rudy LR model because it has Ko as a parameter,
which is essential for our modeling of the ischemic zone.
One deficiency of the space-clamped LR model, which
is carried over from the BeelerReuter model53 from which
it is derived, is that the time-constants for the activation andinactivation of the slow inward Ca current (Is) are an
order of magnitude too large. We have thus decreased the
time constant for the activation of Is(d) by a factor of 10,
which then puts d into the physiologic range.41,54 As in our
prior work on modeling alternans in ischemic muscle,1 we
leave the inactivation time-constant of Is(f) unchanged,
since reducing it results in the level of the plateau of the
action potential being unrealistically depressed. To describe
the other currents we have used the equations appearing in
Table I and the body of the text of Luo and Rudy 1991.44 It
has been noted previously41 that using the equations in Table
I results in currentvoltage relationships for IK1 and IK1( T)that are different from those shown in Figs. 2 and 3B of Luo
and Rudy 1991, respectively. The ionic concentrations
given in Ref. 44 are used to calculate the reversal potentials
ENa , EK , and EK1 we take RT/F26.7 mV for these cal-
culations. In the steady state, the action potential duration
measured between the upstroke and the crossing through of
60 mV on the repolarizing limb of the action potential is
reduced to 237 ms from 290 ms in the standard LR
space-clamped model when paced at a basic cycle length of
400 ms.
We model a one-dimensional strand of ventricular
muscle by the one-dimensional cable equation,55
2V
x 2S Cm VtIion ,
where V is the transmembrane potential mV, x is the spa-
tial coordinate in the strand cm, is the effective longitu-
dinal resistivity (0.2 k cm), S is the surface-to-volume ra-
tio (5000 cm1), Cm is the specific membrane capacitance
(1 F cm2), t is time ms, and Iion is the total ionic current
(A cm2) given by our modified LR model. We use an
explicit integration scheme, with forward Euler integration
for the internal calcium concentration, and an exact analytic
808 Chaos, Vol. 12, No. 3, 2002 Arce, Lopez, and Guevara
Downloaded 12 Sep 2002 to 132.248.28.201. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/chaos/chocr.jsp
-
8/3/2019 Humberto Arce et al- Triggered alternans in an ionic model of ischemic cardiac ventricular muscle
3/12
formula for the activation and inactivation variables for fur-
ther details of the integration scheme, see Refs. 1, 41. The
temporal integration step-size (t) is 0.01 ms and the spatial
integration step-size (x) is 0.01 cm. A lookup table with
linear interpolation (voltage-step0.2 mV) was used to cal-
culate the asymptotic values and the time constants of the
activation and inactivation variables. LHopitals rule was
used where needed to evaluate indeterminate forms.
The one-dimensional space-constant is (Rm /S)1/2
0.06 cm, where Rm is the specific membrane resistance
3.55 k cm2 at the nominal LR value ofKo of 5.4 mM.
The discretization factor (x/) is thus 0.17, at which
point the numerical error is acceptable, as we show below
Fig. 1. The diffusion constant D1/(SC)
103 cm2 ms1, so that the von Neumann linear stability
criterion (x)2/t4D is satisfied. Sealed-end i.e., Di-
richlet boundary conditions are set. Stimulation is carried
out by injecting a 1 ms duration current pulse into the first
five elements of the strand at an amplitude of 100 A cm2
(2-threshold). An implicit integration scheme was also
used to check the results presented below; the findings were
the same, with a small shift in the exact value of Ko atwhich a particular rhythm is seen. Simulations were carried
out on K7 Athlon 650 MHz machines using programs writ-
ten in C 16 significant decimal places.
III. RESULTS
A. Effect of discretization
We first investigate the effect of changing the spacing of
the numerical grid on the propagated action potential in thehomogeneous one-dimensional cable at the nominal Koof 5.4 mM. The cable is allowed to rest for 1000 ms, and
then stimulation is started at a basic cycle length BCL of
400 ms. Figure 1a shows the upstroke of the 21st action
potential at x0.005 cm, 0.01 cm, and 0.025 cm at the
mid-point of a 5 cm long cable with t0.005 ms, 0.01 ms,
and 0.025 ms, respectively. At (x ,t)
(0.025 cm, 0.025 ms), where the discretization factor
(x,) is 0.4, there is an oscillation following the upstroke,
which is much less pronounced at (x ,t)
(0.01 cm, 0.01 ms) and (x,t)(0.005 cm, 0.005 ms).
A similar oscillation has been described previously in the
Beeler Reuter model when the discretization factor is toolarge.56,57 The two upstrokes for (x,t)
(0.01 cm, 0.01 ms) and (x ,t)(0.005 cm, 0.005 ms)
look very similar in Fig. 1a. In fact, Figs. 1b1e show
that several action potential parameters are within a few per-
cent of their asymptotic i.e., x/0 values at (x ,t)
(0.01 cm, 0.01 ms), which agrees with prior work on the
unmodified LR equations.58 For example, decreasing
(x,t) from 0.01 cm, 0.01 ms to 0.005 cm, 0.005 ms
changes the maximum upstroke velocity dV/dtmax or V
max
by 7%, the conduction velocity by 4%, the maximum
voltage overshoot potential in this case (Vmax) by 3%, and
the action potential duration APD by 0.004%. We therefore
use t0.01 ms and x0.01 cm in what follows. The
maximum change in V in a simulation from time t to time
tt is then 2.6 mV.
B. Effect on rhythm of increasing Ko
in theischemic zone
We model the effect of ischemia by increasing Ko in
the central part of the strand the ischemic zone.1,28,37,4043,59
We choose a fixed BCL of 400 ms, which is within the range
used in experimental work on ischemic arrhythmias see Ref.
1 for references. The total length of the strand is 3.0 cm,
with the region of elevated Ko occupying the central 1.0
cm length of the strand, to correspond with our earlier simu-
lations in a two-dimensional sheet.1 At the start of each simu-
lation run at a given Ko , we obtain approximate infinite-
rest initial conditions by setting the variables in the normal-
and high-potassium regions equal to their respective space-
FIG. 1. Action potential propagation in homogeneous one-dimensional
cable atK o5.4 mM. a Action potential upstrokes at middle of a 5 cm
long cable for three different values of the spatial integration step-size ( x)
and the temporal integration step-size (t). x0.005 cm, t
0.005 ms; x0.01 cm, t0.01 ms; x0.025 cm at t0.025 ms.
The 21st action potential after the start of stimulation from infinite-rest
initial conditions stimulus pulse amplitude100 A cm
2; pulse duration1 ms. be Effect of changing t and x on conduction velocity (v),
maximal upstroke velocity dV/dtmax or V
max, maximum potential (Vmax),
and action potential duration APD. The numerical value of t in ms
equals the numerical value of x in cm. v measured between x2.5 cm
and x3.0 cm in the cable; other parameters measured at x2.5 cm.
Curves through the data points are spline fits.
809Chaos, Vol. 12, No. 3, 2002 Triggered alternans
Downloaded 12 Sep 2002 to 132.248.28.201. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/chaos/chocr.jsp
-
8/3/2019 Humberto Arce et al- Triggered alternans in an ionic model of ischemic cardiac ventricular muscle
4/12
clamped steady-state values and then allowing the simulation
to run for 1000 ms, so as to allow some time for equilibration
to occur before injecting the first stimulus at t0 ms. We
now describe the sequence of rhythms seen as Ko is
raised.
Figure 2a left shows action potentials recorded at
three locations x0.75, 1.5, and 2.25 cm in the strand, with
Ko set to its nominal value of 5.4 mM i.e., in the control
situation, before the onset of simulated ischemia. There is a
1:1 response at these and all other sites in the strand, in thateach stimulus produces an action potential of essentially in-
variant morphology neglecting edge effects that propagates
down the entire length of the strand, at a conduction velocity
of 63 cms1. As Ko is raised e.g., Fig. 2a, right:
Ko13.0 mM, the 1:1 response is preserved, but there is
a progressive depolarization of the resting membrane poten-
tial RMP and decrease in the maximum potential ( Vmax)
obtained as the action potential enters the ischemic zone
Fig. 2b. There are also decreases in V max , APD, and
within the ischemic zone see, e.g., Fig. 2a, right: x
1.5 cm. The K-induced decreases in Vmax , V max , and
which have been described previously in ventricular ionic
models,1,28,37,4044,59 are due to a reduction in the fast inward
sodium current (INa) caused by the K-induced depolariza-
tion of the resting potential, while the reduction in APD is
largely due to the K-activation of the maximal conduc-
tances of the outward repolarizing K currents. While the
action potential thus decrements as it enters the ischemic
zone, it increments as it leaves, with the degree of initial
decrement and subsequent increment increasing as Ko is
raised Fig. 2b.
As Ko is increased further, there is eventually a loss
of 1:1 synchronization, with a transition to an alternans or
2:2 rhythm, in which there is a beat-to-beat alternation in themorphology of the action potential between two different
morphologies e.g., Fig. 3a: Ko13.173 mM. The
smaller action potential in Fig. 3a asymptotes to a small
size (amplitude30 mV) towards the distal end of the is-
chemic zone see, e.g., x1.65 cm trace, but then incre-
ments back up to a normal size x2.25 cm trace as it
leaves the ischemic zone. Figure 4 top three traces shows
Vmax as a function of position for the smaller action potential
during alternans at three different values of Ko 13.173,
13.183, and 13.2 mM. This smaller action potential propa-
gates with constant velocity over a considerable part of the
distal ischemic zone, with Vmax staying constant e.g., over a
FIG. 2. Action potential propagation during 1:1 rhythm in the one-
dimensional cable. a Action potentials at three sites: proximal normal zone
top: x0.75 cm, middle of ischemic zone middle: x1.5 cm, distal nor-
mal zone bottom: x2.25 cm. Ko5.4 mM left and 13.0 mM right.
b Resting membrane potential RMP and maximum potential (Vmax) as a
function of distance (x) during a 1:1 rhythm. The ischemic zone lies be-tween x1 cm and x2 cm.
FIG. 3. Action potential propagation during 2:2 and 2:1 rhythms in the
one-dimensional cable. Action potentials at five sites in the one-dimensional
cable during: a 2:2 rhythm Ko13.173 mM, b 2:1 rhythm Ko
13.25 mM.
810 Chaos, Vol. 12, No. 3, 2002 Arce, Lopez, and Guevara
Downloaded 12 Sep 2002 to 132.248.28.201. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/chaos/chocr.jsp
-
8/3/2019 Humberto Arce et al- Triggered alternans in an ionic model of ischemic cardiac ventricular muscle
5/12
distance of 0.25 cm at Ko13.173 mM in Fig. 4.
However, over this part of the ischemic zone, the APD first
decreases slightly and then increases, due to the close spatial
proximity of regions of decremental and incremental conduc-
tion, respectively. When sufficient space for propagation is
available e.g., in a 3 cm homogeneous cable with Ko13.2 mM everywhere, this type of very small-amplitude
action potential asymptotes towards an invariant waveform
similar to that seen at x1.65 cm in Fig. 3a as it proceeds
down the cable. Very similar convergent behavior is also
seen in a homogeneous cable at Ko5.4 mM when Is is
set to zero and INa reduced to a value just above the critical
value that supports a propagating response. This response isthus a genuine propagating response, which we have previ-
ously termed the maintained small-amplitude response. 1
With an increase in Ko , the smaller action potential of
the 2:2 rhythm attains its asymptotic value of Vmax earlier
within the ischemic zone, and the small-amplitude response
recovers back to a full-sized action potential later as it leaves
the ischemic zone see Vmax traces at Ko13.173,
13.183, and 13.2 mM in Fig. 4. Note that at a given Kothe degree of alternation decreases as one moves away from
the center of the ischemic zone and out towards either end of
the cable Fig. 3a.
With a still further increase in Ko , there comes the
point where the smaller action potential does not turn into amaintained small-amplitude response within the ischemic
zone; instead, it decrements sharply within the proximal part
of the ischemic zone, and eventually dies out within the is-
chemic zone Fig. 3b: Ko13.25 mM, resulting in a
2:1 response within the distal portion of the ischemic zone
and throughout the entire distal normal zone. In the part of
the strand proximal to the area of block, there is a 2:2 rhythm
present e.g., Fig. 3b: x0.75, 1.35 cm. As Ko is in-
creased during 2:1 rhythm, the action potential that is even-
tually blocked suffers a greater rate of decrement within the
proximal part of the ischemic zone, so that the site of block
gradually moves to a more proximal location Fig. 4:
Ko13.25, 13.30, and 13.40 mM. A still further increase
in Ko results in period-4, period-6, and period-8 rhythms,
and eventually the appearance of subthreshold rhythms con-
taining no action potentials i.e., complete block at Ko13.6 mM. The sequence of rhythms seen is 1:12:22:14:24:16:26:18:28:12: 01: 0.
C. Bifurcation diagram
One usually constructs a one-parameter bifurcation dia-
gram by plotting the steady-state value of some variable or
the maximum and/or minimum value of that variable as afunction of the bifurcation parameter Ko in our case.
However, the choice of such a variable to act as an index of
the state of the system is not obvious when one has a spa-
tially distributed system such as we have here, since one
would like to have the activity seen at all of the spatial lo-
cations contribute to that index. We thus introduce as our
bifurcation index the normalized voltage integral NVI,
which represents the voltage averaged over space and time
during the course of one 400 ms stimulation cycle. The spa-
tial averaging is obtained by adding together the voltages at
all 300 grid-points at a given time-step. The temporal aver-
aging is then obtained by adding together these sums for
every fifth integration time-step i.e., every 0.05 ms duringeach individual 400 ms stimulation cycle. The resulting sum
of sums is the voltage integral. The normalized voltage inte-
gral is then obtained by dividing this voltage integral by the
voltage integral obtained at Ko5.4 mM.
The bifurcation diagram in Fig. 5, in which we plot NVI
as a function of Ko , summarizes the results described
above, showing the transition from 1:1 to 2:2 rhythm, and
then from 2:2 to 2:1 rhythm, as Ko is increased. The data
points with NVI0.6 during 2:2 rhythm correspond to the
maintained small-amplitude response seen during that
rhythm Figs. 3a, 4. Note also the increase in the value of
the integral for the larger beat upon the transition from 2:2 to
FIG. 4. Maximum potential as a function of distance. Maximum potential
(Vmax) as a function of distance (x) for the smaller of the two responses
during 2:2 rhythm upper three traces: Ko13.173, 13.183, and 13.2
mM and 2:1 rhythm lower three traces: Ko13.25, 13.3, and 13.4
mM. For a 2:1 rhythm lower three traces, when x is sufficiently large,
Vmax equals the resting membrane potential.
FIG. 5. Bifurcation diagram. The normalized voltage integral NVI is plot-
ted as a function of the bifurcation parameter ( K o). The 2:1 rhythm is
seen until Ko13.4 mM.
811Chaos, Vol. 12, No. 3, 2002 Triggered alternans
Downloaded 12 Sep 2002 to 132.248.28.201. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/chaos/chocr.jsp
-
8/3/2019 Humberto Arce et al- Triggered alternans in an ionic model of ischemic cardiac ventricular muscle
6/12
2:1 rhythm, with NVI0.8 corresponding to this propagated
beat during 2:1 rhythm, while NVI0.4 corresponds to the
blocked beat Figs. 3b, 4.
D. Reduction to a one-dimensional map
In clinical,60,61 experimental,3235,6267 and
modeling2224,26,27,42,68 work, it has been shown that analysis
of the response of a cardiac preparation to periodic stimula-tion can often be reduced to consideration of a one- or two-
dimensional finite-difference equation map. One of two
different approaches is usually taken to extract a one-
dimensional map. In the first method, one observes an irregu-
lar rhythm during steady-state conditions and simply plots
some index of the activity on any given beat e.g., APD as a
function of the value of that index on the preceding beat. In
the second method, one applies an S1S2 premature stimula-
tion protocol during 1:1 rhythm, characterizes the premature
beat e.g., by its APD as a function of the S1S2 coupling
interval, and then, using certain assumptions, writes down a
one-dimensional map that can be iterated to then predict the
response of the preparation to periodic driving at any BCL.
Here we use a combination of the above two stimulation
protocols to extract a map. At a given Ko , from infinite-
rest initial conditions, we apply two S1 stimuli at an S1S1
interval of 400 ms, and then inject a single pre- or post-
mature S2 stimulus at a variable S1S2 coupling interval
range 300800 ms. Following the S2 stimulus, we then
apply 30 S3 stimuli at our standard BCL of 400 ms. The
effect of the S1S2 premature stimulation protocol is thus to
change the initial conditions from which periodic pacing at
S3S3400 ms is started. At each S1S2 interval we then cal-
culate the voltage integral for each of the 30 S3 beats fol-
lowing the S2 stimulus, and plot each normalized voltageintegral (NVIi1) as a function of the immediately preceding
integral (NVIi). We shall refer to this map as the voltage
integral map.
Figure 6a illustrates the voltage integral map obtained
at Ko13.0 mM with this S1S2 protocol, where a 1:1
rhythm is seen Fig. 2a, right. A pre- or post-mature S2
stimulus results in a transient episode of alternans, before 1:1
rhythm is re-established asymptotically. This results in the
map having a negative slope. There is a stable period-1 fixed
point on this map at NVI0.78, which, in the simulations,
corresponds to a 1:1 rhythm involving propagation of a
large-sized action potential, as in Fig. 2a right.
There is a gap in the map for 0.33NVIi0.52.The data point just to the left of this gap is produced when
S1S2 is quite short, and corresponds to an action potential
that just barely fails to exit the ischemic zone successfully
(NVIi0.33) being followed by a full-sized action potential
(NVIi10.86). The point just to the right of the gap is
produced at a slightly longer S1S2 interval, when an action
potential that just barely manages to exit the ischemic zone
(NVIi0.52) is followed by a full-sized action potential
(NVIi10.83). An increase in S1S2 of 0.01 ms the small-
est increment we have used is sufficient to convert an action
potential that blocks at the very distal end of the ischemic
zone into one that successfully propagates into and through
the distal normal zone. Thus we have no data points for
0.33
NVIi
0.52.As Ko is increased, the map moves downwards and
to the left, so that the fixed point moves to a lower value; in
the simulations, this corresponds to a fall in NVI as Ko is
increased with 1:1 rhythm being maintained, which is largely
due to the decrease in the amplitude and duration of the
action potential within the ischemic zone Fig. 2b.
As Ko is increased further, a second branch eventu-
ally appears lower down on the map to the right e.g., Fig.
6b: Ko13.16 mM, but there remains a period-1 orbit
corresponding to 1:1 rhythm. With our degree of precision
changing the S1S2 coupling interval in steps as fine as 0.01
ms, the map effectively has a discontinuity between the two
branches at NVIi0.84 in Fig. 6b. Points NVIi ,NVIi1 lying on the right-hand branch in Fig. 6b all have
NVIi0.84 and NVIi10.64 0.67, corresponding to the
fact that Ko is now so high that a sufficiently large action
potential ( NVIi0.84) is now followed by so short a recov-
ery time that the following response is only the maintained
small-amplitude response NVIi10.64 0.67see the
smaller of the two responses of 2:2 rhythm in Fig. 5. In
contrast, should NVIi be sufficiently small (NVIi0.84),
one lands on the left-hand branch, corresponding in the
simulations to a full-sized action potential (NVIi10.76 0.86) being preceded by either a blocked beat
(NVIi0.2 0.4), a maintained small-amplitude response
FIG. 6. One-dimensional maps at 4 different values of K o . Maps pro-
duced by an S1S2 stimulation protocol (S1S2300 800 ms), as describedin the text. a Ko13.0 mM. The map has only one branch. There is a
stable period-1 orbit, corresponding to a 1:1 rhythm. b K o13.16 mM. A second branch of the map can now be seen at the right.
There is still a stable period-1 orbit present, corresponding to 1:1 rhythm.
There is no period-2 orbit. c K o13.173 mM. As well as the stable
period-1 orbit corresponding to 1:1 rhythm, there is also now present astable period-2 orbit arrowheads, corresponding to 2:2 rhythm. d
K o13.34 mM. There is no longer a stable period-1 orbit present.
Rather there is now a stable period-2 orbit arrowheads, corresponding to
2:1 rhythm.
812 Chaos, Vol. 12, No. 3, 2002 Arce, Lopez, and Guevara
Downloaded 12 Sep 2002 to 132.248.28.201. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/chaos/chocr.jsp
-
8/3/2019 Humberto Arce et al- Triggered alternans in an ionic model of ischemic cardiac ventricular muscle
7/12
(NVIi0.6 0.7), or a full-sized action potential (NVIi0.75 0.84). The slope of the right-hand branch of the map
is also negative, for a reason similar to that outlined above
for the left-hand branch: i.e., a larger action potential (NVI i)
on this branch will be followed by a smaller maintained
small-amplitude response (NVIi1). As Ko is increased,
the right-hand branch moves down and to the left, so that the
point of effective discontinuity in the map also moves to
the left.At Ko13.173 mM, the map admits a stable
period-2 orbit, as well as the period-1 orbit just described
above Fig. 6c. This period-2 orbit corresponds to an al-
ternans rhythm consisting of a maintained small-amplitude
response (NVI0.65) alternating with a full-sized action po-
tential (NVI0.82). Each of the period-1 and period-2 or-
bits has its own basin of attraction. As Ko is slightly
increased, the period-1 orbit disappears, leaving only the
period-2 orbit corresponding to 2:2 rhythm.
For sufficiently high Ko , the right-hand branch of the
map is found to lie much lower down and there is a period-2
orbit involving this branch Fig. 6d: Ko13.34 mM.
This period-2 orbit corresponds to a 2:1 rhythm consisting ofa full-sized action potential (NVI0.86) alternating with a
blocked beat (NVI0.37)see Figs. 4 and 5. This very
abrupt transition from 2:2 rhythm to 2:1 rhythm occurs when
the incremental propagation that turns the maintained small-
amplitude response into a full-sized action potential during
2:2 rhythm Figs. 3a, 4 becomes decremental conduction
upon exiting the ischemic zone. The abruptness of this tran-
sition can be appreciated from the fact that decreasing the
coupling interval by only 0.01 ms in an S1S2 premature
stimulation protocol suffices to convert an incrementing into
a decrementing response at Ko13.0 mM, producing ef-
fectively all-or-none propagation.
E. 1:1^2: 2 bistability
In all of the numerical simulation results presented
above Figs. 2 5, each simulation run was started from only
one set of initial conditions. However, the map analysis
above predicts that bistability should occur Fig. 6c. Fig-
ure 7 shows that dropping a single stimulus from the periodic
drive train at that exact value of Ko 13.173 mM does
indeed result in the conversion of the 1:1 rhythm into a 2:2
rhythm, demonstrating the existence of 1:12:2 bistabil-ity. In addition, the reverse transition from 2:2 to 1:1 rhythm
can be induced by the injection of a suitably timed extra-
stimulus.
F. Piecewise-linear map
Our simulations Figs. 25, 7 and the NVI maps Fig.
6 indicate that the sequence of rhythms 1: 1 alone1: 1
coexisting with 2:22: 2 alone2:1 is seen as Ko is
increased. We decided to explore the bifurcation sequence
systematically using a piecewise-linear map that is an ap-
proximation to the maps in Figs. 6a 6c. For Ko suf-
ficiently low, the map has a single branch, which is the
straight line NVIi10.2(NVIi)0.9 Fig. 8a, which
has a negative slope that is 1 in absolute value. There is
therefore a globally-attracting period-1 orbit on this map,
corresponding to 1:1 rhythm compare Fig. 8a with Fig.
6a. As Ko is increased, a second branch, with the same
slope, makes its appearance Fig. 8b. The equation of this
branch is NVIi10.2(NVIi)0.8, so that there is a jump
discontinuity of size 0.1 between the two branches of the
map compare Fig. 8b with Fig. 6b. Nevertheless, the
period-1 orbit remains globally attracting. An increase in
Ko corresponds to gradually moving the position of the
jump discontinuity at NVIi to the left, leaving the size
of the jump unchanged see also Figs. 6b and 6c. There
eventually comes a point where a period-2 orbit appears
corresponding to 2:2 rhythm, which coexists with the
period-1 orbit compare Fig. 8c with Fig. 6c. With a still
further increase in Ko , and resultant movement to the left
in the position of the jump discontinuity i.e., decrease in ,
the left-hand branch of the map eventually no longer inter-
sects the line of identity, so that the period-1 orbit disappears
in a discontinuous fashion, leaving only the period-2 orbitbehind Fig. 8d. Figure 9 gives the bifurcation diagram of
the piecewise-linear map, with the bifurcation parameter be-
ing the location of the discontinuity in the map. As
decreases corresponding to an increase of Ko in the
cable simulations, one sees the transition from a period-1
orbit 1:1 rhythm alone to coexisting period-1 and period-2
orbits bistable 1:1 and 2:2 rhythms and then to a period-2
orbit 2:2 rhythm alone.
G. Rhythms in space-clamped membrane
It is obvious that the various rhythms described above in
the cable are largely determined by the properties of the
FIG. 7. Conversion from 1:1 to 2:2 rhythm caused by dropping one stimulus
from the periodic drive train. The 1:1 rhythm existed for 100 cycles before
the stimulus was dropped, and the 2:2 rhythm persisted for 100 stimuli thelength of the longest integration run that we made following the resumption
of stimulation after the stimulus was dropped.
813Chaos, Vol. 12, No. 3, 2002 Triggered alternans
Downloaded 12 Sep 2002 to 132.248.28.201. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/chaos/chocr.jsp
-
8/3/2019 Humberto Arce et al- Triggered alternans in an ionic model of ischemic cardiac ventricular muscle
8/12
K-depolarized membrane within the central ischemic zone.
A question that naturally arises is whether the response of the
distributed system i.e., the one-dimensional cable to peri-
odic stimulation can then be accounted for by the response of
space-clamped K
-depolarized membrane to periodic stimu-lation. However, this is a very difficult question to investi-
gate systematically for two reasons. First, the waveform of
the stimulation current entering any particular isopotential
grid-element of the cable is very complicated in shape, since
the axial or longitudinal current is proportional to V/x.
Second, this current changes as a function of position along
an inhomogeneous cable. For example, during a 2:1 block in
the cable, a grid-element lying sufficiently proximal within
the ischemic zone will show a local 2:2 response, while an
identical grid-element lying sufficiently distal will show a
local 2:1 response Figs. 3b, 4. This difference in responsemust thus be due entirely to the fact that the net longitudinal
current source minus sink is different in the two grid-
elements.
In an attempt to investigate this problem, we study the
response of the space-clamped membrane to periodic deliv-
ery of a 1 ms duration current-pulse stimulus at a fixed BCL
of 400 ms over a wide range of Ko and pulse amplitude.
At each fixed value of the stimulus amplitude 27 A cm2
threshold at the nominal Ko of 5.4 mM to
54 A cm2, incremented in steps of 1 A cm2, Ko is
changed in the range 520 mM. At each pulse amplitude, a
1:1 rhythm is seen when Ko is sufficiently low, and more
complex rhythms can appear as Ko is increased. In par-ticular, a transition from 1:1 to 2:2 rhythm can occur, and
bistability between these two rhythms can also be seen: e.g.,
at a stimulus current of 30 A cm2, which is just above
threshold, and Ko14.6677 mM, the 1:1 rhythm that ap-
pears can be converted into a 2:2 rhythm by dropping a
stimulus pulse, producing a waveform very similar to that
shown at x15 mm in Fig. 7. The similarity of the rhythms
and waveforms seen in the space-clamped and distributed
cases thus reinforces the conclusion that the local properties
of the membrane within the ischemic zone play a key role in
determining the overall behavior in the cable.
IV. DISCUSSION
A. The transition from 1:1 to 2:1 rhythm
A transition from 1:1 to 2:1 rhythm is often seen in
experimental17,18,30,35,6367,6971 and modeling22,67,72 work on
ventricular muscle and Purkinje fiber as the pacing frequency
is increased or when some intervention is made that de-
creases the effective stimulus amplitude or the excitability of
the tissue, e.g., ischemia,3,6 elevation of Ko .73,74 This
transition to 2:1 rhythm can be direct or indirect. For ex-
ample, in paced isolated rabbit ventricular cells, the transi-tion is direct when the stimulus amplitude is intermediate in
size.67 In contrast, when the stimulus amplitude is higher or
lower, alternans17 or Wenckebach63,71 rhythms are seen, re-
spectively, before the 2:1 rhythm occurs. Similar results are
found in aggregates of driven spontaneously beating embry-
onic chick ventricular cells.64 It remains to be seen in our
model whether changing some parameter other than Koe.g., the length of the ischemic zone, the inter-cellular cou-
pling within the ischemic zone,72 or the basic cycle length72,
might result in a direct transition to 2:1 rhythm or an indirect
transition via Wenckebach rhythms as Ko is then in-
creased. It also remains to be seen whether a transition from
FIG. 8. Piecewise-linear maps approximating the NVI maps from the cable
simulations (Fig. 6). a There is a globally attracting period-1 orbit, corre-sponding to 1:1 rhythm in the simulations. b the position of the discon-
tinuity in the map0.9. Although a second branch now exists on the map,there is still a globally attracting period-1 orbit, corresponding to 1:1
rhythm. c 0.76. There are now stable period-1 and period-2 orbits,
corresponding to 1:1 and 2:2 rhythms, respectively. d 0.70. Theperiod-1 orbit has disappeared, so that the only stable orbit now present is a
period-2 orbit, corresponding to a 2:2 rhythm.
FIG. 9. Bifurcation diagram for the piecewise-linear map. The bifurcation
parameter is the position of the discontinuity in the maps of Fig. 8. Note
that there is a range of over which there is bistability between 1:1 and 2:2
rhythms. The increment in is 0.01. At each value of, 100 iterations were
carried out from each of 100 equally-spaced initial conditions. The first 96
iterates were not plotted in each case, to allow time for any transients to
pass.
814 Chaos, Vol. 12, No. 3, 2002 Arce, Lopez, and Guevara
Downloaded 12 Sep 2002 to 132.248.28.201. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/chaos/chocr.jsp
-
8/3/2019 Humberto Arce et al- Triggered alternans in an ionic model of ischemic cardiac ventricular muscle
9/12
2:2 to 4:4 rhythm, and not to 2:1 rhythm, might occur, as has
been reported in fast driving of toad ventricle 75 and Purkinje
fiber.35,68
In the cable simulations we report on above, the transi-
tion is indirect, since alternans is seen before 2:1 rhythm
occurs Figs. 3 5. In thin pieces of ventricular endocardium
uniformly exposed to high Ko , alternans is also seen be-
fore a 2:1 response occurs as Ko is elevated in the bath-
ing solution.74
The alternans in that case involves two full-sized action potentials, as occurs within both normal zones in
our case e.g., Fig. 3a: x0.75 cm and x2.25 cm. How-
ever, the alternans within the ischemic zone here involves the
maintained small-amplitude response e.g., Fig. 3a: x
1.65 cm.
B. Ischemic alternans
Alternans is quite frequently seen in the electrocardio-
gram or electrogram in the acute stage of myocardial is-
chemia, in both clinical and experimental work see the ref-
erences given in the Introduction. There are also recordings
of the transmembrane potential showing a beat-to-beat alter-
nation in the action potential morphology during ischemia
e.g., Refs. 3, 6, 8. Indeed, the 1:12:22:1 progres-sion we see above in the distal half of the cable as Ko is
raised is exactly what is seen within the first few minutes of
ischemia, with the reverse sequence being seen after the
coronary occlusion is then released see, e.g., Fig. 3 of Ref.
3.
The exact mechanisms underlying the ischemic alternans
seen with any of the above three modes of recording remain
unknown. Nevertheless, it is of interest to note that regional
hyperkalemia induces alternans and ventricular
arrhythmias.
76
It is not known whether the alternation seen inthe action potential morphology is primary i.e., intrinsic
to the cell or whether it is secondary i.e., due to electro-
tonic coupling with neighboring regions showing a 2:1
response.1,72 One can make the case for both of these forms
of alternans in our model. Primary alternans is seen through-
out the cable when there is a 2:2 rhythm e.g., Fig. 3a. In
addition, a 2:2 rhythm closely resembling that seen within
the ischemic zone in the cable simulations is seen in our
simulations of space-clamped K-depolarized membrane.
However, the fact that the range of Ko over which the 2:2
rhythm exists in the cable is so narrow only 0.05 mM
wide in Fig. 5 leads one to the conclusion that primary
alternans might be quite rare during ischemia. Alternans isalso seen in the cable during 2:1 block over a much wider
range of Ko (0.2 mM); it then occurs throughout the
entire proximal normal-Ko segment e.g., Fig. 3b: x
0.75 cm and within the proximal part of the high-Koischemic central zone Fig. 3b: x1.35 cm. Alternans in
the normal-Ko segment of tissue has also been seen in a
Purkinje fiber strand where Ko is elevated along a central
8 mm long segment to produce a 2:1 block.73
One interesting finding here is the small-amplitude de-
flection seen during 2:2 rhythm e.g., at x1.65 cm in Fig.
3a. When such a small deflection is recorded experimen-
tally at one site, the assumption is usually made that this is a
subthreshold response that is in the distal part of a pathway
displaying decremental conduction. However, our simula-
tions show that this is not necessarily the case, since this
response can in fact be a nondecrementing response.
C. Triggered alternans
Our simulations show that alternans can be triggered by
a pause in stimulation during 1:1 rhythm Fig. 7 and detrig-gered by an extrastimulus. As pointed out earlier Introduc-
tion, the sudden appearance or disappearance of alternans
following a premature beat has been noted during experi-
ments on ischemia several times.2,4,8,9 It is not completely
clear from these studies whether it is the premature beat itself
or the subsequent compensatory pause that is responsible for
the flip from one rhythm to the other. The latter prospect is
supported by the fact that a deliberate pause in stimulation
can result in the triggering of alternans,8 which is reminis-
cent of the pause-induced triggering shown in Fig. 7. How-
ever, as already mentioned, the range of Ko over which
the 2:2 rhythm is bistable with the 1:1 rhythm in our model
is so narrow 0.05 mM in Fig. 5, that one would not ex-
pect to stay within the bistable range of Ko for a very
long time during rapidly evolving acute myocardial is-
chemia. It is of course possible that the range in the model is
too narrow because of some deficiency in the model e.g., the
simple elevation ofKo might not be enough to accurately
represent ischemia.28,38,39,42,59
In another experiment,4 the premature beat preceding
triggering of alternans was quite late, so that the compensa-
tory pause preceding triggering was not very long. Should
the pause indeed be the cause of the triggering in that case,
such a short pause would almost certainly not be long
enough to produce paused-induced triggering of alternans inour model. Whereas detriggering of alternans can occur fol-
lowing a spontaneous premature beat and compensatory
pause2,4,8 or an intentional pause in stimulation,8 a pause
during alternans does not detrigger the alternans in our
model. This suggests that a different mechanism might be
involved in detriggering than in triggering, or even that the
mechanism of pause-induced triggering seen in our model
might have little to do with the mechanism of pause-induced
triggering during ischemia.
More than a century ago, Gaskell suggested that one way
to account for mechanical alternans would be if certain por-
tions of the ventricle respond only to every second impulse,
while other portions respond to every stimulus.77
One in-triguing possibility is thus that the triggering of alternans in
the experiments might be due to the induction of a region of
2:1 block which would then result in secondary alternans,
since it is known that a pause can result in a flip from 1:1 to
2:1 rhythm in the intact ventricle69 and in isolated rabbit
ventricular cells,67 as can an injection of a suitably timed
extrastimulus in isolated ventricular cells.67 An intentionally
inserted extrastimulus can also convert 2:1 into 1:1 rhythm,
provided that the timing is right,67,69 suggesting a mechanism
that might be capable of producing detriggering of secondary
alternans. However, a pause in this circumstance would
again not be expected to detrigger the secondary alternans,
815Chaos, Vol. 12, No. 3, 2002 Triggered alternans
Downloaded 12 Sep 2002 to 132.248.28.201. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/chaos/chocr.jsp
-
8/3/2019 Humberto Arce et al- Triggered alternans in an ionic model of ischemic cardiac ventricular muscle
10/12
since a pause does not convert 2:1 to 1:1 rhythm in isolated
rabbit ventricular cells.67 One can even speculate, should it
be possible to flip back from alternans to 1:1 rhythm, that
this maneuver might remove the well-known proarrhythmic
effect of alternans. Further systematic experimental work ex-
ploring the origin of triggering and detriggering of alternans
during ischemia is thus indicated.
A flip between high- and low-amplitude ischemic alter-
nans following an extrasystole has been described.5,6,8
Thereis also evidence for this bistability indeed, perhaps multista-
bility or neutral stability during mechanical alternans of the
nonischemic ventricle.7880 Thus, by our definition of trig-
gered alternans, it is important to rule out the possibility that
there is a pre-existing very-low-amplitude alternans present
see, e.g., Fig. 3 of Ref. 2; Fig. 5a of Ref. 8 before stating
that triggered alternans occurs. There is evidence that the flip
from a lower-amplitude to a higher-amplitude alternans dur-
ing ischemia is due to the conversion of discordant alternans
alternation 180 out-of-phase at some sites into concordant
alternans alternation in-phase at all sites.5,6 While there
have been modeling studies showing discordant alternans in
nonischemic models,2931 it remains to be seen whether dis-cordant alternans will be seen in models of ischemic muscle,
and whether flips between the concordant and discordant
forms of alternans will be possible. Such flips can be seen in
a homogeneous cable model29 and in an elegantly simple
phenomenological model.19
D. Bistability involving 1:1, 2:2, and 2:1 rhythms
Four forms of bistability involving 1:1, 2:2, and 2:1
rhythms have been described: 2: 22: 1, 1:12:1 ,1:12:2 , and 2:2A2:2B i.e., bistability between
two different 2:2 rhythms.First, pulse- or pause-induced flips can be seen between
2:2 and 2:1 rhythms in isolated rabbit ventricular cells,17
hysteresis between these two rhythms suggesting the exis-
tence of bistability has been described in bullfrog
ventricle,66 and either a 2:2 or a 2:1 rhythm can be seen in a
homogeneous 1-dimensional Beeler Reuter cable, depend-
ing on initial conditions.27 The existence of the 2:22:1 bistability is predicted by a simple one-dimensional two-
branched discontinuous map derived from the APD restitu-
tion curve, and involves the co-existence of a stable period-1
orbit on one branch of the map with a stable period-doubled
period-2 orbit on the other branch of the map.27,33
Second, 1:12:1 bistability is found in frogventricle,69 in aggregates of embryonic chick ventricular
cells,64 in single rabbit ventricular cells,67 in the human His
Purkinje system,81 and in the space-clamped LuoRudy
model.67 Hysteresis between these two rhythms has also
been described.24,64,66,67,69 The 1:12:1 bistability is pre-dicted by a simple one-dimensional two-branched map de-
rived from the APD restitution curve,27,33,67 and involves the
co-existence of two different period-1 points on the two
branches of the discontinuous map27,33,67 but see Ref. 66 for
a dissenting view.
Third, we show an example of1:12:2 bistability inFig. 7. Apart from the triggered ischemic alternans men-
tioned above, the only other two instances of 1:12: 2bistability of which we are aware occur in the mechanical
contraction of the heart7880 and in periodic driving of modi-
fied forms of the space-clamped BeelerReuter
equations.2224 In the latter case, the behavior is different
from ours, in that the 2:2 cycle is made up of two full-sized
action potentials compare, e.g., the inset in Fig. 4b of Ref.
23 with our Fig. 3a. An iterative analysis32 was carried out
for that case using the APD restitution curve.22
This analysiscan be transformed into consideration of a one-dimensional
(APDi ,APDi1) map having a single branch of negative
slope,27,33 which is in contrast to our effectively two-
branched map Fig. 8c. In a differently modified version of
the Beeler Reuter model, the iterative approach predicted
that there would be bistability between 1:1 and 2:2
rhythms,23,24,26 due to the coexistence of stable period-1 and
period-2 orbits on the single branch of the map when the
Schwarzian derivative is negative for a unimodal map, there
can be at most one stable periodic orbit.82 An unstable
period-2 orbit then serves to separate the basins of attraction
of the stable period-1 and the stable period-2 orbits, which is
not the case in Fig. 6c, where the discontinuity and itspre-image serve that purpose. In the ionic modeling work,
the unstable period-2 orbit arises from a subcritical period-
doubling bifurcation that occurs as the bifurcation parameter
the basic cycle length is decreased.23,26 Very subtle changes
in the APD restitution curve can result in the subcritical bi-
furcation becoming supercritical, thus destroying the un-
stable period-2 orbit and the bistability.26 As in our case Fig.
5, the bistability found in the numerical simulations existed
only over an extremely narrow range of the bifurcation pa-
rameter as little as 1 ms of basic cycle length.23,24 This
bistability in the space-clamped system has been linked with
the bistability between periodic and quasiperiodic modes of
sustained reentry around a one-dimensional loop.24 We know
of no experimental report in single cells of this kind of bi-
stability.
Fourth, bistability has been described between two dif-
ferent 2:2 rhythms in frog ventricle,32 and a bistability or
multistability or neutral stability between 2:2 rhythms has
been described in the mechanical activity of the dog
heart.7880 This bistability has also been treated in terms of a
single-branched one-dimensional map.27 In addition, as de-
tailed above, concordant and discordant alternans can be
viewed as coexisting 2:2 rhythms.
E. Nonischemic alternans; alternans control
While alternans can be a precursor of ventricular ar-
rhythmias in situations not involving ischemia e.g., rapid
pacing,83 long-QT syndrome,15 administration of anti-
arrhythmic drugs84, we are not aware of any reports of trig-
gered alternans in these nonischemic cases. Our expectation
is that triggered alternans will show up in these and other
situations in which alternans is know to occur, once a sys-
tematic search is made.
Alternans in the atrioventricular conduction time, in-
duced in human beings by a pacing protocol, can be made to
revert to a 1:1 atrioventricular rhythm by using a modifica-
816 Chaos, Vol. 12, No. 3, 2002 Arce, Lopez, and Guevara
Downloaded 12 Sep 2002 to 132.248.28.201. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/chaos/chocr.jsp
-
8/3/2019 Humberto Arce et al- Triggered alternans in an ionic model of ischemic cardiac ventricular muscle
11/12
tion of a chaos-control technique, which involves reduc-
tion of the dynamics to consideration of a one-dimensional
map.85 The excellent approximation of the dynamics de-
scribed above by a one-dimensional map makes it more fea-
sible that ischemic alternans, a well-known precursor of ma-
lignant arrhythmias, can be made to revert back to a less
ominous period-1 rhythm by means of some sort of altern-
ans control technique.
ACKNOWLEDGMENTS
We thank CONACYT Grant No. 32164-E H.A., A.L.,
PAPIME-UNAM Grant No. 191054 H.A., and the Cana-
dian Institutes of Health Research M.R.G. for financial sup-
port, and Enrique Palacios-Boneta for general help with
computers.
1 H. Arce, A. Xu, H. Gonzalez, and M. R. Guevara, Alternans and higher-
order rhythms in an ionic model of a sheet of ischemic ventricular
muscle, Chaos 10, 411426 2000.2 H. K. Hellerstein and I. M. Liebow, Electrical alternation in experimental
coronary artery occlusion, Am. J. Physiol. 160, 366374 1950.3 E. Downar, M. J. Janse, and D. Durrer, The effect of acute coronary
artery occlusion on subepicardial transmembrane potentials in the intact
porcine heart, Circulation 56, 217224 1977.4 K. Kataoka and R. Yoshimura, Clinical and experimental studies on elec-
trical alternans of ST-segment and T-wave in variant form of angina pec-
toris in Japanese, Resp. Circ. 27, 767776 1979.5 H. Hashimoto, K. Suzuki, and M. Nakashima, Effects of the ventricular
premature beat on the alternation of the repolarization phase in ischemic
myocardium during acute coronary occlusion in dogs, J. Electrocardiol.
17, 229238 1984.6 H. Hashimoto, M. Asano, and M. Nakashima, Potentiating effects of a
ventricular premature beat on the alternation of the ST-T complex of epi-
cardial electrograms and the incidence of ventricular arrhythmias during
acute coronary occlusion in dogs, J. Electrocardiol. 17, 289301 1984.7 D. L. Carson, R. Cardinal, P. Savard, and M. Vermeulen, Characterisa-
tion of unipolar waveform alternation in acutely ischaemic porcine myo-
cardium, Cardiovasc. Res. 20, 521527 1986.8 S. G. Dilly and M. J. Lab, Electrophysiological alternans and restitution
during acute regional ischaemia in myocardium of anaesthetized pig, J.
Physiol. London 402, 315333 1988.9 R. Mohabir, M. R. Franz, and W. T. Clusin, In vivo electrophysiological
detection of myocardial ischemia through monophasic action potential re-
cording, Prog. Cardiovasc. Dis. 34, 1528 1991.10 D. R. Adam, S. Akselrod, and R. J. Cohen, Estimation of ventricular
vulnerability to fibrillation through T-wave time series analysis, Comput-
ers in Cardiology IEEE, New York, 1982, pp. 307310.11 D. R. Adam, J. M. Smith, S. Akselrod, S. Nyberg, A. O. Powell, and R. J.
Cohen, Fluctuations in T-wave morphology and susceptibility to ven-
tricular fibrillation, J. Electrocardiol. 17, 209218 1984.12 S. Horinaka, S. Hara, N. Tsuchiya, A. Yabe, H. Asakawa, H. Yagi, Y. Mori,
and H. Matsuoka, Alternans of ventricular gradient during percutaneous
transluminal coronary angioplasty, J. Electrocardiol. 34, 135139
2001.13 B. D. Nearing, A. H. Huang, and R. L. Verrier, Dynamic tracking of
cardiac vulnerability by complex demodulation of the T wave, Science
252, 437440 1991.14 P. Laguna, G. B. Moody, J. Garcia, A. L. Goldberger, and R. G. Mark,
Analysis of the ST-T complex of the electrocardiogram using the
KarhunenLoeve transform: adaptive monitoring and alternans detec-
tion, Med. Biol. Eng. Comput. 37, 175189 1999.15 S. B. Platt, J. M. Vijgen, P. Albrecht, G. F. van Hare, M. D. Carlson, and
D. S. Rosenbaum, Occult T wave alternans in long QT syndrome, J.
Cardiovasc. Electrophysiol. 7, 144148 1996.16 S. H. Hohnloser, T. Klingenheben, Y.-G. Li, M. Zabel, J. Peetermans, and
R. J. Cohen, T wave alternans as a predictor of recurrent ventricular
tachyarrhythmias in ICD recipients: prospective comparison with conven-
tional risk markers, J. Cardiovasc. Electrophysiol. 9, 12581268 1998.17 M. R. Guevara, F. Alonso, D. Jeandupeux, and A. C. G. van Ginneken,
Alternans in periodically stimulated isolated ventricular myocytes: Ex-
periment and model, in Cell to Cell Signalling: From Experiments to
Theoretical Models, edited by A. Goldbeter Harcourt Brace Jovanovich,
London, 1989, pp. 551563.18 J. Hescheler and R. Speicher, Regular and chaotic behaviour of cardiac
cells stimulated at frequencies between 2 and 20 Hz, Eur. Biophys. J. 17,
273280 1989.19 D. S. Rubenstein and S. L. Lipsius, Premature beats elicit a phase rever-
sal of mechanoelectrical alternans in cat ventricular myocytes: A possible
mechanism for reentrant arrhythmias, Circulation 91, 201214 1995.20
E. Chudin, J. Goldhaber, A. Garfinkel, J. Weiss, and B. Kogan, Intracel-lular Ca2 dynamics and the stability of ventricular tachycardia, Bio-
phys. J. 77, 29302941 1999.21 J. Huser, Y. G. Wang, K. A. Sheehan, F. Cifuentes, S. L. Lipsius, and L. A.
Blatter, Functional coupling between glycolysis and activation
contraction coupling underlies alternans in cat heart cells, J. Physiol.
London 524, 795806 2000.22 A. Vinet, D. R. Chialvo, D. C. Michaels, and J. Jalife, Nonlinear dynam-
ics of rate-dependent activation in models of single cardiac cells, Circ.
Res. 67, 15101524 1990.23 A. Vinet and F. A. Roberge, Analysis of an iterative difference equation
model of the cardiac cell membrane, J. Theor. Biol. 170, 201214
1994.24 A. Vinet, Memory and bistability in a one-dimensional loop of model
cardiac cells, J. Biol. Syst. 7, 451473 1999.25 J. J. Fox, J. L. McHarg, and R. F. Gilmour, Jr., Ionic mechanism of
electrical alternans, Am. J. Physiol. 282, H516H530 2001.26 A. Vinet, Quasiperiodic circus movement in a loop model of cardiac
tissue: multistability and low dimensional equivalence, Ann. Biomed.
Eng. 28, 704720 2000.27 T. J. Lewis and M. R. Guevara, Chaotic dynamics in an ionic model of
the propagated cardiac action potential, J. Theor. Biol. 146, 407432
1990.28 J. M. Ferrero, Jr., B. Rodriguez, B. Trenor et al., Action potential dura-
tion inhomogeneities in acute myocardial ischemia: a simulation study,
Computers in Cardiology IEEE, New York, 1998, pp. 577580.29 M. A. Watanabe, F. H. Fenton, S. J. Evans, H. M. Hastings, and A. Karma,
Mechanisms for discordant alternans, J. Cardiovasc. Electrophysiol. 12,
196206 2001.30 J. J. Fox, M. L. Riccio, F. Hua, E. Bodenschatz, and R. F. Gilmour, Jr.,
Spatiotemporal transition to conduction block in canine ventricle, Circ.
Res. 90, 289296 2002.
31 Z. Qu, A. Garfinkel, P.-S. Chen, and J. N. Weiss, Mechanisms of discor-dant alternans and induction of reentry in simulated cardiac tissue, Cir-
culation 102, 16641670 2000.32 J. B. Nolasco and R. W. Dahlen, A graphic method for the study of
alternation in cardiac action potentials, J. Appl. Physiol. 25, 191196
1968.33 M. R. Guevara, G. Ward, A. Shrier, and L. Glass, Electrical alternans and
period-doubling bifurcations, Computers in Cardiology IEEE Computer
Society, Silver Springs, MD, 1984, pp. 167170.34 M. R. Guevara and A. Shrier, Rhythms produced by high-amplitude
periodic stimulation of spontaneously beating aggregates of embryonic
chick ventricular myocytes, Ann. N.Y. Acad. Sci. 591, 1122 1990.35 D. R. Chialvo, R. F. Gilmour, Jr., and J. Jalife, Low dimensional chaos in
cardiac tissue, Nature London 343, 653657 1990.36 J.-M. Cao, Z. Qu, Y.-H. Kim, T.-J. Wu, A. Garfinkel, J. N. Weiss, H. S.
Karagueuzian, and P.-S. Chen, Spatiotemporal heterogeneity in the in-
duction of ventricular fibrillation by rapid pacing. Importance of cardiacrestitution properties, Circ. Res. 84, 13181331 1999.
37 B. Tilg and P. Wach, Simulation of the 1-D cardiac excitation process by
varying extracellular potassium concentration and intracellular resistiv-
ity, Computers in Cardiology IEEE, New York, 1995, pp. 421424.38 J. M. Ferrero, Jr., V. Torres, J. Saiz, M. Monserrat, J. M. Ferrero, and N. V.
Thakor, Simulation study of a ction potentials during acute myocardial
ischemia, Comput. Cardiol. 24, 705708 1997.39 R. M. Shaw and Y. Rudy, Electrophysiologic effects of acute myocardial
ischemia. A mechanistic investigation of action potential conduction and
conduction failure, Circ. Res. 80, 124 138 1997.40 A. Cimponeriu, C. F. Starmer, and A. Bezerianos, Action potential
propagation in ischemic cardiac tissue: a theoretical computer modeling,
in Ref. 28, pp. 317320.41 A. Xu and M. R. Guevara, Two forms of spiral-wave reentry in an ionic
model of ischemic ventricular myocardium, Chaos 8, 157174 1998.
817Chaos, Vol. 12, No. 3, 2002 Triggered alternans
Downloaded 12 Sep 2002 to 132.248.28.201. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/chaos/chocr.jsp
-
8/3/2019 Humberto Arce et al- Triggered alternans in an ionic model of ischemic cardiac ventricular muscle
12/12
42 A. B. Feldman, H. D. Esperer, and R. J. Cohen, A mechanism controlling
the characteristics of T-wave alternans measurements in the setting of
myocardial ischemia, Comput. Cardiol. 26, 149152 1999.43 Y. Wang and Y. Rudy, Action potential propagation in inhomogeneous
cardiac tissue: safety factor considerations and ionic mechanism, Am. J.
Physiol. 278, H1019H1029 2000.44 C.-h. Luo and Y. Rudy, A model of the ventricular cardiac action poten-
tial. Depolarization, repolarization, and their interaction, Circ. Res. 68,
15011526 1991.45 S. Guan, Q. Lu, and K. Huang, A discussion about the DiFrancesco
Noble model, J. Theor. Biol. 189, 2732 1997.46 A. Varghese and G. R. Sell, A conservation principle and its effect on the
formulation of NaCa exchanger current in cardiac cells, J. Theor. Biol.
189, 3340 1997.47 L. P. Endresen, K. Hall, J. S. Hye, and J. Myrheim, A theory for the
membrane potential of living cells, Eur. Biophys. J. 29, 90103 2000.48 L. P. Endresen and N. Skarland, Limit cycle oscillations in pacemaker
cells, IEEE Trans. Biomed. Eng. 47, 11341137 2000.49 S. Dokos, B. G. Celler, and N. H. Lovell, Modification of DiFrancesco
Noble equations to simulate the effects of vagal stimulation on in vivo
mammalian sinoatrial node electrical activity, Ann. Biomed. Eng. 21,
321335 1993.50 A. Varghese and R. L. Winslow, Dynamics of the calcium subsystem in
cardiac Purkinje fibers, Physica D 68, 364386 1993.51 A. Varghese and R. L. Winslow, Dynamics of abnormal pacemaking
activity in cardiac Purkinje fibers, J. Theor. Biol. 168, 407420 1994.52
T. J. Hund, J. P. Kucera, N. F. Otani, and Y. Rudy, Ionic charge conser-vation and long-term steady state in the LuoRudy dynamic cell model,
Biophys. J. 81, 33243331 2001.53 G. W. Beeler and H. Reuter, Reconstruction of the action potential of
ventricular myocardial fibres, J. Physiol. London 268, 177210 1977.54 G. Isenberg and U. Klockner, Calcium currents of isolated bovine ven-
tricular myocytes are fast and of large amplitude, Pfluegers Arch. 395,
3041 1982.55 G. H. Sharp and R. W. Joyner, Simulated propagation of cardiac action
potentials, Biophys. J. 31, 403423 1980.56 L. Boucher and J.-P. Drouhard, Analyse par simulation des effets de la
discretisation sur le potential daction ventriculaire propage. Institut de
Genie Biomedical, Universite de Montreal, Montreal. Technical Report,
1983.57 B. Victorri, F. A. Roberge, L. Boucher, A. Vinet, and J. P. Drouhard, in
Ref. 33, pp. 431434.58
J. Wu and D. P. Zipes, Effects of spatial segmentation in the continuousmodel of excitation propagation in cardiac muscle, J. Cardiovasc. Elec-
trophysiol. 10, 965972 1999.59 R. M. Shaw and Y. Rudy, Electrophysiologic effects of acute myocardial
ischemia: a theoretical study of altered cell excitability and action poten-
tial duration, Cardiovasc. Res. 35, 256272 1997.60 W. Mobitz, Uber die unvollstandige Storung der Erregungsuberleitung
zwischen Vorhof und Kammer des menschlichen Herzens, Z. Gesamte
Exp. Med. 41, 180237 1924.61 A. Shrier, H. Dubarsky, M. Rosengarten, M. R. Guevara, S. Nattel, and L.
Glass, Prediction of complex atrioventricular conduction rhythms in hu-
mans with use of the atrioventricular nodal recovery curve, Circulation
76, 11961205 1987.62 M. R. Guevara, L. Glass, and A. Shrier, Phase locking, period-doubling
bifurcations, and irregular dynamics in periodically stimulated cardiac
cells, Science 214, 13501353 1981.63 M. R. Guevara, D. Jeandupeux, F. Alonso, and N. Morissette, Wenck-
ebach rhythms in isolated ventricular heart cells, in International Con-
ference on Singular Behavior and Nonlinear Dynamics, Vol. 2, edited by
St. Pnevmatikos, T. Bountis, and Sp. Pnevmatikos World Scientific, Sin-
gapore, 1989, pp. 629642.64 M. R. Guevara, A. Shrier, and L. Glass, Chaotic and complex cardiac
rhythms, in Cardiac Electrophysiology: From Cell to Bedside, 1st ed.,
edited by D. P. Zipes and J. Jalife Saunders, Philadelphia, 1990, pp.
192201.
65 M. L. Koller, M. L. Riccio, and R. F. Gilmour, Jr., Dynamic restitution of
action potential duration during electrical alternans and ventricular fibril-
lation, Am. J. Physiol. 275, H1635H1642 1998.66 G. M. Hall, S. Bahar, and D. J. Gauthier, Prevalence of rate-dependent
behaviors in cardiac muscle, Phys. Rev. Lett. 82, 29952998 1999.67 A. R. Yehia, D. Jeandupeux, F. Alonso, and M. R. Guevara, Hysteresis
and bistability in the direct transition from 1:1 to 2:1 rhythm in periodi-
cally driven single ventricular cells, Chaos 9, 916931 1999.68 R. F. Gilmour, Jr., N. F. Otani, and M. A. Watanabe, Memory and com-
plex dynamics in cardiac Purkinje fibers, Am. J. Physiol. 272, H1826
H1832 1997.69 G. R. Mines, On dynamic equilibrium in the heart, J. Physiol. London
46, 349383 1913.70 D. R. Chialvo and J. Jalife, Non-linear dynamics of cardiac excitation
and impulse propagation, Nature London 330, 749752 1987.71 A. R. Yehia, A. Shrier, K. C.-L. Lo, and M. R. Guevara, Transient out-
ward current contributes to Wenckebach-like rhythms in isolated rabbit
ventricular cells, Am. J. Physiol. 273, H1H11 1997.72 M. R. Guevara, Spatiotemporal patterns of block in an ionic model of
cardiac Purkinje fibre, in From Chemical to Biological Organization,
edited by M. Markus, S. C. Muller, and G. Nicolis Springer-Verlag, Ber-
lin, 1988, pp. 273281.73 P. F. Cranefield, H. O. Klein, and B. F. Hoffman, Conduction of the
cardiac impulse. 1. Delay, block, and one-way block in depressed Purkinje
fibers, Circ. Res. 28, 199219 1971.74 M. L. Koller, M. L. Riccio, and R. F. Gilmour, Jr., Effects of K
oon
electrical restitution and activation dynamics during ventricular fibrilla-
tion, Am. J. Physiol. 279, H2665H2672 2000.75 G. V. Savino, L. Romanelli, D. L. Gonzalez, O. Piro, and M. E. Val-
entinuzzi, Evidence for chaotic behavior in driven ventricles, Biophys.
J. 56, 273280 1989.76 M. J. Curtis, The rabbit dual coronary perfusion model: a new method
for assessing the pathological relevance of individual products of the is-
chaemic milieu: role of potassium in arrhythmogenesis, Cardiovasc. Res.
25, 10101022 1991.77 W. H. Gaskell, On the rhythm of the heart of the frog, and on the nature
of the action of the vagus nerve, Philos. Trans. R. Soc. London, Ser. A
173, 9931033 1882.78 W. G. Guntheroth, B. C. Morgan, G. A. McGough, and A. M. Scher,
Alternate deletion and potentiation as the cause of pulsus alternans,
Am. Heart J. 78, 669681 1969.
79 Y. Mahler and S. Rogel, Interrelation between restitution time-constantand alternating myocardial contractility in dogs, Clin. Sci. 39, 625639
1970.80 S. Suzuki, J. Araki, Y. Doi, W. Fujinaka, H. Minami, G. Iribe, S. Mohri, J.
Shimizu, M. Hirakawa, and H. Suga, Coupling interval from slow to
tachycardic pacing decides sustained alternans pattern, Am. J. Physiol.
280, H1368H1375 2001.81 E. Pruvot, A. De Torrente, G. M. De Ferrari, P. J. Schwartz, and J.-J. Goy,
Two-to-one AV block associated with the congenital long QT syn-
drome, J. Cardiovasc. Electrophysiol. 10, 108113 1999.82 G. Mayer-Kress and H. Haken, Transition to chaos for maps with posi-
tive Schwarzian derivative, in Evolution of Order and Chaos, edited by
H. Haken Springer-Verlag, Berlin, 1982, pp. 183186.83 J. M. Pastore, S. D. Girouard, K. R. Laurita, F. G. Akar, and D. S. Rosen-
baum, Mechanism linking T-wave alternans to the genesis of cardiac
fibrillation, Circulation 99, 13851394 1999.84 H. Tachibana, M. Yamaki, I. Kubota, T. Watanabe, S. Yamauchi, and H.
Tomoike, Intracoronary flecainide induces ST alternans and reentrant
arrhythmia on intact canine heart. A role of 4-aminopyridine-sensitive
current, Circulation 99, 16371643 1999.85 D. J. Christini, K. M. Stein, S. M. Markowitz, S. Mittal, D. J. Slotwiner,
M. A. Scheiner, S. Iwai, and B. B. Lerman, Nonlinear-dynamical ar-
rhythmia control in humans, Proc. Natl. Acad. Sci. U.S.A. 98, 5827
5832 2001.
818 Chaos, Vol. 12, No. 3, 2002 Arce, Lopez, and Guevara